Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Propagation of Improvement Effects of Critical Inter-city Link --- The Japanese Decadal Change of the Available Travel Routes ---

Makoto OKUMURA Makoto TSUKAI Professor Associate Professor Center for Northeast Asian Studies Graduate School of Engineering Tohoku University Hiroshima University 41, Kawauchi, Aoba-ku, Sendai 1-4-1, Kagamiyama, Higashi-Hiroshima 980-8576 Japan 739-8527 Japan Fax: +81-22-795-7477 Fax: +81-82-424-7827 E-mail: [email protected] E-mail: [email protected]

Yusuke KIMURA East Japan Railway Company Operation and Rolling Stock Department 2-2-2, Yoyogi, Shibuya-ku, Tokyo 151-8578 Japan

Abstract: Unlike the urban highway network, the inter-city network is composed of heterogeneous links of various speeds, costs and frequencies. Improvement of one airway link can possibly provide a new attractive travel route, such as an air connection route, as well as multi-modal route, including middle distance railway access. The improvement effect of a critical link may be propagated over the nation-wide network, unlimited to the direct neighbor connector of the improved link. How such propagation occurred in Japan between 1995 and 2005 is investigated. K-th shortest path search algorithm is applied to find the available set of routes. Remarkable changes in the Tohoku-Western Japan pairs are detected. This paper further analyzes the effects of the Shinkansen expansion to Hachinohe in 2002 on the service level of the available routes for those OD pairs, proving that there are strong multi-modal propagations on the air links at Sendai .

Key Words: inter-city network, multi-modal routes, set of routes, Shinkansen

1. INTRODUCTION

1.1 Background Financing schemes for transportation development have been drastically diversified in recent decades. While European countries having long history of transportation operations that separate infrastructure provision from operations, developing countries try to take BOT, BOO and internalize the operation profit as to cover the initial building cost. When we design an appropriate financial scheme, we must analyze the flow of the investment effect, but such analysis for an inter-city transportation project is not so easy as that for an urban transportation project, where most users can be limited to the inhabitants of that city, especially along the line of the project. That condition is not the case in large scale inter- urban transportation projects, but transportation economists insist that the benefit of the improvement is generated in the transport service market at first, then, construction cost should be covered with the overcharge on the fare, and be gathered from the operators. Based on that belief, cost allocation scheme for new Shinkansen (rapid train system) Project is controversial, especially how they determine the cost sharing among the Japanese national government, local governments along the lines, and JR, as an operator. For example, JR West Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009 is asked to cover a part of the construction cost of the Kyushu Shinkansen to be operated by JR Kyushu, while JR East is asked to cover a part of the cost of the Hokuriku Shinkansen expanded into the area of JR West.

Furthermore, such effect propagation would not be limited to one mode --- railways in the above examples. As we will show in this paper, a newly constructed Shinkansen line can be used as a middle-distance access to a local airport, and a route shift is occurring to support an air link departing at the local airport. Such wide, and multi-modal propagation is a unique phenomenon in an inter-city network already constructed to certain degree of density.

1.2 Related Studies In the inter-city passenger transportation, competition between railway and air service have been frequently discussed and analyzed (Janic, 2003; Gonzalez-Savignat, 2004). Recently, however, complementally mixture of these two different modes gathers interests: one of the typical mixture use is railway access service for an airport (Lythgoe and Wardman, 2002), which take advantages of reliability of the railway service. Railway service has also another characteristics that it can collect demands distributed thinly along the line. It can be used as an effective demand collector for a hub airport, especially when it is difficult to expand the airport for several reasons (Givoni and Banister, 2006).

In reality, there are not a few Multi-modal routes are chosen by passengers (Horn, 2003; Tsukai and Okumura, 2003), but the connectivity between the different modes is far from the satisfactory (O’Sullivan and Patel, 2004; Tapiador et al., 2009). From the passengers’ viewpoint, not only physical and spatial, but also temporal connectivity are very crucial (Cascetta and Papola, 2003; Krygsman et al., 2004; Malighettia et al., 2008). Concerning the measurement index for connectivity, Daily Reachable Sphare and Maximum Stay Duration are proposed in the Japanese national infrastructure development planning process (Sato and Totani, 2005; Morichi et al., 2005). Mathematical methodologies to find several routes on a dense network, and to efficiently calculate the indices have been studied (Kato et al., 1978; Miller-Hooks and Patterson, 2004). The present authors applied these methods to the Japanese inter-city public transportation network (Tsukai and Okumura, 2005; 2006), followed with the propose of a network design model (Okumura and Tsukai, 2007). However, the realistic service level evaluations have not yet executed, mainly because of data availability problem.

1.3 The Aim of This Paper This paper aims to show how such propagation occurred in Japan between 1995 and 2005. The k-th shortest path search algorithm is applied to find the available set of inter-city trip routes. Remarkable changes in Tohoku-Western Japan pairs, due to the Tohoku Shinkansen expansion up to Hachinohe City in 2002, are detected. This paper analyzes the change of the service level of the available routes in those OD pairs, by comparing the air, multi-modal, and rail routes from the Hachinohe zone to the zones in the Western. It is proved that there are strong multi-modal propagations of the Shinkansen expansion on the air links at , owing to the improvement of frequency and maximum staying time at the destination.

Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

2. CHANGE OF THE AVAILABLE SET OF TRAVEL ROUTES

2.1 Network Data An inter-regional network database was prepared corresponding with the spatial unit of the Japanese Inter-city Net Passenger Traffic Survey using 207 zone divisions. We used 194 areas for the analysis excluding some isolated islands. The dataset consists of 246 nodes (194 zone centroids and 52 airport nodes) with 560 links (217 air, 291 trunk railway, and 52 airport access links). The prepared attributes of each link were physical distance, travel time, standard fare (for air and access links), name of express train, and the hourly frequencies of 18 hours from 6:00 to 24:00 for each direction. The source of the diagram information is referred to the timetables published in October 1995 and in October 2005.

Major changes of inter-city transportation services around this decade are summarized in table 1. Besides the two international of Kansai and Chubu, most of the airport improvements appeared in rural area - far from the metropolitan regions (events 3,4,9 in table 1). This is why the improvements of the frequency was not so large as that of the travel time, especially in the case of new airports in rural areas, where flight frequency to (Tokyo) could not so increase as to have been expected, due to the capacity limit of Haneda. The Hachinohe expansion of the Tohoku Shinkansen in 2002 (event 8) cut 34 minutes of nominal travel time, and the previous transfer time at Morioka Station became unnecessary, while the increase of frequency was small. The Kyushu Shinkansen solely did not have any direct service to Tokyo in the Shinkansen projects, shown in table 1 (event 11), but it provided large amount of shortening of travel time, as well as the remarkable increase of frequency as much as 16.5 times in a day. While the opening of Chubu International Airport resulted in an increase in the frequency of air ways (event 12), Kansai International Airport has lost more than half of its air lines by 2005.

Table 1 Major events of Japanese inter-city transportation around 1995-2005 travel time, frequency, no. of lines No. year date event 1995.10 2005.10 unit 1992 3.14 Nozomi (270km/h) on Tokaido Shinkansen (Tokyo-Osaka) 16 53.5 times 7.18 Yamagata Mini-Shinkansen (Fukushima-Yamagata) opened 11.5 14 times 1993 3.18 Nozomi on Tokaido-Sanyo Shinkansen (Tokyo-Fukuoka) 13.5 17 times 1994 9.4 Kansai International Airport (Osaka) opened 68 / 25 45 / 12 times / lines 1 1997 3.22 Akita Mini-Shinkansen (Morioka-Akita) opened 104 / 13.5 84 / 14 minutes / times 2 1998 4.5 Akashi Strait Bridge (Hyogo) opened with bus service 3 7.18 Odate-Noshiro (North Akita) Airport (Akita) opened 3 / 2 times / lines 4 7.28 (Saga) opened 5 / 2 times / lines 5 10.1 Nagano Shinkansen (Takasaki-Nagano) opened 91 / 18.5 51 / 18 minutes / times 6 1999 12.4 Yamagata Mini-Shinkansen (Yamagata-Shinjo) expanded 53 / 9 46 / 8.5 minutes / times 7 2000 3.23 New B Runway of Haneda Airport (Tokyo) replaced 250 / 36 373 / 40 times / lines 8 2002 12.1 Tohoku Shinkansen (Morioka-Hachinohe) expanded 71 / 14 37 / 16 minutes / times 9 2003 7.7 Noto Airport (Ishikawa) opened 2 / 1 times / lines 10 10.1 Shinagawa Station on Tokaido Shinkansen (Tokyo) opened 16 54 times 11 2004 3.13 Kyushu Shinkansen (Yatusushiro-Kagoshima) opened 122 / 12 39 / 28.5 minutes / times 12 2005 2.17 Chubu International Airport (Nagoya) added to old airport 61 / 25 106 / 29 times / lines

Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Hokkaido

Aomori AP. Misawa AP. ③ Hachinohe ⑧ ① Zone ⑥ Hokuriku Tohoku Chugoku ⑨ Sendai AP. ⑤ ⑩ Kanto Kyushu ⑦ ② ⑫ Tokai ④ Kinki ⑪ 0 500km Shikoku Figure 1 Location of the Major Projects and Division of Regions

2.2 Route Search Algorithm The available travel route for each OD pair is defined as follows: At first, the shortest time paths in 1995 and 2005 are found by the Dijkstra algorithm, without considering transfer time or frequency. Secondly, k-th path search algorithm is used to search other routes up to 40. If travel time of that route exceeds 1.5 times of the shortest time in 1995, or if that route includes more than three air links, that route is deleted form the available set. Comparing to a sparse air network, the railway network is too dense to produce many similar routes. Therefore, several calculated routes including the common air links, are represented by the shortest time route of them. This procedure means that all routes including only railway links are represented by the shortest one.

The k-th path search algorithm, proposed by Kato, et. al (1978) consists of the Dijkstra algorithm, second path search routine (FSP) including Dijkstra, and k-th path search algorithm (KSP) including FSP and Dijkstra. This algorithm can generate the shortest path with a sequential order from the shortest to k-th shortest path between an arbitral pair of origin and destination nodes. See Tsukai and Okumura (2005) to find the details about the procedure. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Table 2 Average number of available travel routes per one OD pair classeified by region Region Hokkaido Tohoku Kanto Tokai Hokuriku Kinki Chugoku Shikoku Kyushu 11.1 13.9 16.1 10.1 14.8 11.1 7.7 8.8 Hokkaido 11.9 15.6 17.5 9.8 14.8 12.5 8.1 9.7 0.8 9.4 8.7 13.1 9.7 14.1 15.2 12.0 Tohoku +6.9% 8.9 9.5 12.3 10.4 16.0 14.5 14.6 1.7 -0.5 6.0 5.9 9.7 11.8 11.2 13.9 Kanto +11.2% -5.1% 7.4 5.8 8.8 11.7 9.8 15.3 1.4 0.7 1.5 3.6 2.6 7.3 6.3 13.0 Tokai +8.2% +6.9% +20.5% 4.3 2.0 6.4 5.9 13.4 -0.3 -0.8 -0.1 0.7 5.4 9.6 10.2 8.7 Hokuriku -2.5% -6.7% -1.2% +1.3% 6.2 9.7 9.3 9.8 0.1 0.7 -0.9 -0.5 0.9 6.5 5.9 13.0 Kinki +0.4% +7.0% -8.4% -11.6% +10.5% 5.0 4.9 11.3 1.4 1.9 0.0 -0.9 -0.1 -1.4 8.4 8.7 Chugoku +11.9% +12.4% -0.2% -12.2% -0.8% -24.1% 6.3 8.4 0.4 -0.6 -1.4 -0.4 -0.8 -0.9 -2.1 9.1 Shikoku +4.6% -4.0% -12.3% -6.7% -8.6% -25.2% -25.7% 7.9 0.9 2.6 1.4 0.4 1.1 -1.8 -0.3 -1.2 Kyushu +9.5% +19.8% +9.6% +2.7% +11.0% -12.7% -7.4% -12.5% Note: No. of routes in 1995 No. of routes in 2005 Increase (2005-1995) Change Rate(%) (05-95)/95 2.3 Number of Available Travel Routes The result of the route search is summarized for the nine regions as table 2. The upper-right triangle shows the average number of available routes in 1995 and 2005 for one OD pair included in the region pair, while the bottom-left triangle shows the difference and ratio of them. At a glance, the numbers are larger in the remote regions, compared to the near regions. The reason of this result comes from the representation of the multi-modal routes including the common air links and representation of railway routes. The short distance OD pairs are usually connected with a railway route, and hardly have other routes up to 1.5 times of the shortest time, including air links. Besides that, there are fewer routes for the Shikoku Region than for other cases.

2.4 Change of the Available Routes

2.4.1 Effect of Airport Openings As shown in table 1, newly opened airports excluding Chubu International, are located in rural areas and the new air service is limited to small destinations such as Haneda or Itami. Therefore, the opening of Odate-Noshiro Airport (Akita) (event 3) never contributed to increase the routes between Tohoku and Western Japan (Kinki, Chugoku and Kyushu). Similarly, the opening of Saga Airport (event 4) did not affect on the number of routes between Kyushu and the Eastern Japan (from Hokkaido to Hokuriku). While the increase of Eastern Japan lines from Chubu International (event 12) gave positive effects on the route availability between Tokai region to Eastern Japan, but not the case for Western regions. In Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Table 3 Average number of available routes from Hachinohe zone Average No. of Routes Kinki Chugoku Shikoku Kyushu 1995 8.1 10.3 10.9 8.0 Hachinohe 2005 10.9 18.8 15.8 16.3 Increase (2005-1995) 2.8 8.6 4.9 8.4 Change Rate(%) (05-95)/95 34% 84% 44% 105%

Table 4 Service levels of air lines at Hachinohe zone Airport Year Misawa Aomori Sendai 1995 45 108 159 Access time (min) 2005 55 99 120 Frequency to 1995 1 2 7 Osaka (Itami) 2005 1 2 13 1995 0 1+4/7 2 Frequency to Kansai 2005 0 1 0 1995 0 3/7 2 Frequency to Fukuoka 2005 0 4/7 5 Frequency to other West Japan 1995 0 0 2 (3cities) (Hiroshima,Okayama,Takamatsu) 2005 0 0 3 (3cities) Number of daily users 1995 66529 96276 14620 2005 53885 51445 18114 (increase) 05-95 -12644 -44831 3494 Note: 3/7 means 3 times per week all, we can conclude that there are not strong positive effects of airport openings on the inter- city route availability.

2.4.2 Effect of Air link Service Change Table 2 shows a remarkable increase of travel routes from the Hokkaido area. This result is because of the opening of eight air links inside the Hokkaido region, especially the air services from which yield new air connection routes from other regions.

In the Hokuriku region, the number of available routes to other regions has decreased, despite the opening of the Noto Airport (Ishikawa)(event 9). This is the result of the abandonment of the flight service between Toyama and Nagoya, which concurrently occurred in the same decade. The opening of the Akashi Strait Bridge connecting the Kinki region and the Shikoku Island in 1999 (event 2) encouraged new frequent express bus services which resulted in the abandonment of the competing air service between Itami (Osaka) – Tokushima, Itami – Takamatsu (Kagawa), Kansai – Takamatsu, Kansai – Kochi, and Nagoya – Takamatsu. The failure of commuter air service between Hiroshima and Matsuyama (Ehime) also affected negatively on the route availability between Shikoku and other Chugoku, Kinki, Kanto regions.

We should mention that the domestic hub role of Kansai International Airport was strongly decreased due to the wide abandonment of air service for Tohoku, Chugoku and Kyushu Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009 airports, as well as Shikoku airports.

2.4.3 Effect of Openings of Shinkansen Although the opening of the Shinkansen service may cause the abandonment of competing air services (such as Akita – Itami in 2003) in the middle run, we can detect a positive effect on the number of available routes by the calculation in this section, excluding the positive effect on alternate rail routes. Faster Shinkansen service accessing to alternative airports in different areas, which sometimes have more air lines than the local airport, increased the number of available travel routes.

We can show examples of this effect in all new Shinkansen lines. The most particular one is the Hachinohe zone connected by the Tohoku Shinkansen expansion (event 8). New multi- modal routes are added by using Sendai airport for 120 minutes of access time, rather than 159 minutes before. Table 3 shows the remarkable increase in the average number of the available routes for OD pairs from the Hachinohe zone to zones in the Western Japan regions. This increase of routes is a result of the Shinkansen access improvement to Sendai, Fukushika and Haneda airports, in addition to the traditionally used local airports such as Misawa, Aomori and Hanamaki, through the Shinkansen service. The difference in frequencies for those airports is shown in table 4. The number of routes using Sendai airport increased from 35 to 227, that using from 0 to 65, and that using Haneda airport from 0 to 47. We will analyze the service level change more precisely in the next section.

In addition, the Nagono Shinkansen (event 5) provides new routes including Haneda Airport to Nagano and Ueda zones. The Kyushu Shinkansen (event 11) provides accessibility of Kumamoto and Fukuoka Airports to Kagoshima and Sendai zone, as well as accessibility of to Kumamoto zone, mutually. Those multi-modal effects are reflected in table 2.

3. CHANGE OF THE SERVICE LEVELS OF TRAVEL ROUTES

3.1 Estimation of the Service Levels for each Available Route In this section, the change of the service levels for the available routes from the Hachinohe zone to zones in Western Japan, are examined in more detail. To accomplish this objective, we first estimate several practical service levels under the user’s point of view, based on the attributes data of each link in year of 1995 and 2005.

3.1.1 Estimation of Route Distance and Nominal Time k k Route distance ( D od ) and nominal travel time (T od ) are simply calculated as a sum of physical distance and travel time of the links included in the route.

Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

3.1.2 Estimation of Monetary Cost k Monetary cost of route ( C od ) is estimated by the summation of the cost of air and access k k (CA od ) and the cost of railway ( CR od ). The former is given simply as the summation of link fare. On the other hand, the latter railway cost is estimated by the following manner, in order to consider the difference of express fare related to the travel speed, and non-linearity to k k the travel distance. Firstly, railway distance ( DR od ) and railway nominal travel time ( TR od ), then nominal railway speed is calculated. k k k vR od = DR od /TR od × 60 (1) Based on the nominal speed, we select one of the following formulas originally estimated by statistical analysis, in order to describe the express fare, = k + δ = k ≤ F1ex .3 46DR od 972.37, ( 1 :1 vR od 120) , (2) = k + δ = < k ≤ F 2ex .5 70DR od 1971 ,0. ( 2 :1 120 vR od 170) , (3) = k + δ = < k F3ex .6 49DR od 2528 ,5. ( 3 :1 170 vR od ) . (4) On the other hand, normal nonlinear railway fare of JR can be well estimated by the following quadratic function. = + k − k 2 Fn 58.47 16.49DR od .0 0043(DR od ) (5) As a result, monetary cost of railway in a route is given as follows. k = + δ + δ + δ CR od Fn 1F1ex 2F 2ex 3F3ex (6)

3.1.3 Estimation of Effective Frequency In our earlier papers (Tsukai and Okumura, 2003; 2005), we gave the minimum value of daily link frequencies included in the route as the daily route frequency, but that method clearly overestimates the effective frequency, because it neglects the time passage. Tsukai and Okumura (2006) proposed a more precise procedure to follow the time passage based on the link frequencies per every three hours. The present study takes that procedure into account to utilize hourly link frequency data.

The procedure is based on the idea that a traveler’s trip schedule must be selected from the available trains or flights. At first we assume in each link, trains (flights) are set with the same interval time in each hour. If the hourly frequency of an hour beginning h o’clock is f h , the i -th train (flight) is assumed to depart at the node at 2( i − 2/)1 f h × 60 minutes after h o’clock. Secondly, we begin the simulation of the time passage at the origin node, and select one starting train (flight) of the first link. We add the link nominal travel time and proceed to the following node, until we meet the different train (flight) name in the following link. If we meet a different name, waiting time before the transfer is considered. We select the earliest departure time of the next link no earlier than the arrival time of the node. When the simulation reaches the destination node, the arrival time of that starting time is determined. Thirdly, such a calculation is repeated for each of the all trains (flights) at the first link departing from the origin node. If the calculated time exceeds 24:00, that starting time is judged as unreachable. If the arrival becomes the same time with that of the later starting train (flight), such schedule is considered to be ineffective.

Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Table 5 Service level comparison over the representative routes

1995 2005 Change (2005-1995) Rail Air M.M. M.M.-Air Rail Air M.M. M.M.-Air Rail Air M.M. M.M.-Air Distance (km ） 1195 1098 1077 -21 1206 1115 1029 -85 12 17 -48 -64 Kinki Nominal Travel Time (min. ） 449 216 278 62 365 224 261 37 -85 9 -17 -25 Monetary Cost (yen ） 22519 36670 37115 445 23895 41670 40813 -856 1376 5000 3699 -1301 Effective Frequency 7.55 1.09 1.95 0.87 9.23 1.18 7.93 6.75 1.68 0.09 5.98 5.89 22 No. of Reacheable zones 22 21 22 1 22 21 22 1 0 0 0 0 zones EAT at destination (min) 15:27 17:05 15:44 -81 13:39 17:03 12:02 -301 -108 -1 -221 -220 LDT at destination (min) 13:43 13:13 14:14 60 16:25 13:46 16:15 149 162 32 121 88 EAT at Hachinohe (min) 15:29 15:23 15:00 -22 12:58 14:54 12:44 -131 -151 -29 -137 -108 LDT at Hachinohe (min) 14:18 11:38 12:44 66 16:35 11:01 17:32 391 137 -37 288 325 Zones of positive MSD 10 1.5 8 6.5 19.5 1.5 19.5 18.0 10 0 11.5 12 Average of positive MSD (min) 38 135 82 231 165 330 Distance (km ） 1585 1317 1468 151 1586 1321 1302 -19 1 5 -165 -170 Chugoku Nominal Travel Time (min. ） 566 243 281 38 473 252 304 52 -94 9 23 14 Monetary Cost (yen ） 27079 46970 50389 3419 28155 55467 51160 -4307 1076 8497 770 -7727 Effective Frequency 3.99 1.33 1.97 0.64 6.37 1.41 5.47 4.06 2.37 0.08 3.50 3.42 19 No. of Reacheable zones 17.5 17.5 19 1.5 19 17 19 2 2 -0.5 0 0.5 zones EAT at destination (min) 18:07 16:34 16:06 -28 16:08 15:53 12:56 -178 -118 -41 -190 -149 LDT at destination (min) 11:44 13:20 13:09 -11 13:56 12:47 14:23 97 133 -33 74 107 EAT at Hachinohe (min) 18:04 15:37 16:11 34 15:34 14:37 14:19 -18 -150 -60 -112 -52 LDT at Hachinohe (min) 11:33 11:01 12:07 65 14:39 11:57 15:59 242 185 56 233 177 Zones of positive MSD 0 1 1 0 6 0.5 12.5 12.0 6 0 12.0 12 Average of positive MSD (min) 20 11 128 62 226 Distance (km ） 1537 1216 1326 110 1548 1241 1243 2 11 25 -83 -108 Shikoku Nominal Travel Time (min. ） 595 242 305 63 510 256 311 55 -85 15 7 -8 Monetary Cost (yen ） 26075 42143 47516 5374 27566 50893 48864 -2029 1492 8750 1348 -7402 Effective Frequency 3.96 1.00 1.79 0.79 6.04 1.36 5.89 4.54 2.07 0.36 4.11 3.75 14 No. of Reacheable zones 14 14 14 0 14 14 14 0 0 0 0 0 zones EAT at destination (min) 18:31 19:52 17:51 -120 16:29 17:36 13:29 -248 -122 -135 -263 -127 LDT at destination (min) 10:42 13:39 13:58 20 13:36 13:11 15:04 113 174 -27 65 93 EAT at Hachinohe (min) 18:09 14:54 15:32 38 15:40 14:22 13:47 -35 -149 -32 -105 -73 LDT at Hachinohe (min) 11:11 8:32 10:39 127 13:54 10:25 16:10 346 163 113 332 219 Zones of positive MSD 0 0 2 0 3.5 0.5 11 10.5 4 1 9.0 11 Average of positive MSD (min) 78 47 48 169 Distance (km ） 1953 1513 1633 120 1963 1551 1539 -12 11 38 -94 -132 Kyushu Nominal Travel Time (min. ） 706 277 334 57 597 285 339 54 -109 8 5 -3 Monetary Cost (yen ） 29994 48245 52490 4246 30888 56266 55138 -1128 894 8022 2648 -5374 Effective Frequency 3.14 0.78 1.36 0.58 4.61 0.75 4.75 4.00 1.47 -0.03 3.39 3.43 26 No. of Reacheable zones 22.5 23 23.5 0.5 25.5 22.5 26 3.5 3 -0.5 2.5 3 zones EAT at destination (min) 18:58 19:52 17:32 -140 17:41 19:43 14:13 -330 -77 -9 -199 -190 LDT at destination (min) 10:18 13:38 11:57 -101 12:01 13:21 14:24 63 103 -17 147 164 EAT at Hachinohe (min) 19:33 14:52 15:55 63 17:25 14:21 14:33 12 -128 -31 -82 -51 LDT at Hachinohe (min) 10:20 8:55 10:12 77 12:19 7:42 15:15 453 119 -73 303 376 Zones of positive MSD 0 0 1 1.0 2 0 16.5 16.5 2 0 15.5 16 Average of positive MSD (min) 64 30 153 In this way, we can count the number of reachable and effective starting time at the origin node for the given travel route. If the result numbers for each direction are different, we define the average of those two values as the effective frequency of that travel route.

3.1.4 Earliest Arrival Time, Latest Departure Time to Return, and Maximum Stay Duration Through the above procedure of effective frequency calculation, we can get the earliest arrival time at the destination under the condition that we can start at the origin no earlier than 6:00. Similarly, we can determine the latest time of departure of return trip at the destination node, in order to arrive back at the origin node no later than 24:00. Those two values are used for the earliest arrival time (EAT) and the latest departure time to return (LDT), and the interval between them gives the maximum stay duration at the destination (MSD). Those three indices are direction dependent. Then we use the directional average of those values for evaluation.

Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

3.2 Service Level Comparison over the Representative Route of the Different Mode

3.2.1 Definition of the Representative Route of the Different Mode In order to assess the role of railways and air for the route service level, we compare the three different type of mode, with each other; air, rail and multi-modal. Comparison will be done over the representative route from each category. The railway route is uniquely calculated in this research as mentioned in section 2.2.

At first, we determine the nearest airport from the origin node and that from the destination node. Air routes are defined as the available routes that use one or more air links connecting between the nearest airports. Furthermore, those air routes are represented by the route having the shortest nominal travel time. Multi-modal routes are defined as the available routes that use air link, but either of the nearest airports is not used. In other words, this type includes middle distance railway access, as well as an air link. Furthermore, we select a representative one in this category based on the effective frequency.

3.2.2 Service Level Comparison Comparison of service level indices over the representative routes is summarized in table 5. Both in 1995 and 2005, the difference of travel distance and nominal travel time is similar for air routes and multi-modal routes, while they are much shorter than those for the railway route. On the other hand, the monetary cost shows the opposite order: the railway route is much cheaper than the other two including the more expensive air service. In 1995, the cost of an air route is slightly cheaper than that of a multi-modal, but that is not the case in 2005. Rail routes have superiority of effective frequency over multi-modal routes in both years, with the exception of Kyushu in 2005, while air routes are always lag behind, in terms of effective frequency.

In table 5, EAT and LDT indices are shown for each direction for the travel from Hachinohe to Western Japan, or in the opposite direction. Although the nominal travel time of air routes are short, EAT at the Kinki, Shikoku and Kyushu destinations of air route are not earlier than that of the rail route, as well as the multi-modal route. On the other hand, EAT at Hachinohe of an air route is comparable to that of a multi-modal route, and much earlier than that of a rail route. This difference comes from the flight schedule on the without night stays, where the first flight departure is late in the morning. The very low frequency of Misawa airport also makes that the LDT of an air route at Hachinohe as early as 8:00 in the morning. As a result, the air route cannot take its fast speed advantage and fails to provide a positive MSD for most of the zones in Western Japan.

On the contrary, the multi-modal routes take much advantage of high frequencies and provide much earlier EAT and later LDT than air routes in many cases, especially in 2005. As a result, the multi-modal routes successfully provide positive MSD in many destinations, while both air and rail almost fails for the zones in Chugoku, Shikoku and Kyushu regions. Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

3.3 Decadal Difference of the Service Levels

3.3.1 Change of Railway Route Service Introduction of rapid and frequent service of the Shinkansen has given the multiple positive effects on the service levels of the railway routes. The 37 minutes reduction of link travel time between Hachinohe and Morioka (see table 1) is reinforced with other railway speed up projects to provide 85 through 109 minutes reduction of the nominal travel time to Western Japan. The effective frequencies were improved around 1.5 in Kyushu, and 2.4 in Chugoku. These combinational changes yielded the EAT more than two hours earlier than before, and the LDT two or three hours later. Such remarkable improvement in terms of the time dimension resulted in the increase of the zones that have positive MSD in 2005. Because we indirectly estimate the railway cost based on a regression formula, the speed improvement is concurrent with the cost increase. But the change of the cost does not seem serious: it is as small as 6 percent change.

3.3.2 Change of Air Route Service The service level of air routes did not improve very much, although the airfare was raised from 5,000 to 9,000 yen in the decade. There were not large effects from either improvement of Western lines from Chubu or decline of lines from Kansai, because both in 1995 and 2005, no air service was provided either from Misawa or . As a result, the number of zones reachable in a day and that of positive MSD was not augmented.

3.3.3 Change of Multi-modal Route Service Multi-modal routes improved considerably, owing to the usability of several airports of more frequent service. Because the length of the air way in route distance is not so long, and limited on the trunk lines where air carrier was reluctant to raise the airfare considering the fierce competition to Shinkansen service, the increase of the monetary cost was suppressed as small as 3,700 yen at most. Large positive impacts are certified in all destination regions: effective frequency increased between 3.4 to 6.0 per day, EAT became earlier for 82 to 260 minutes, LDT got later for one to three hours, and the remarkable increase of the zones of positive MSD.

It is very important that those service improvements in multi-modal routes are usually larger than that in railway routes. We can say that the impact of the Shinkansen expansion was propagated with multi-modality. The far right column of table 5 shows that the superiority of multi-modal routes to the air routes were remarkably expanded in the decades. It implies that we must pay much attention to multi-modal connectivity in the nation-wide inter-city transportation planning.

3.4. Evidence in the Passenger Behavior

Improvement of multi-modal routes including air link from Sendai airport is widely accepted by the users, and resulted in the increase of such routes in 2005. Evidence of this is the increase of Sendai airport users in the travelers between Hachinohe and zones in Western Japan region, already shown in table 4, with a remarkable contract to the decrease of users of Misawa and Aomori airports. Figure 2 also shows another evidence: the weekday users of Sendai airport remarkably increased in the zones including the expanded Shinkansen, such as Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

Number of daily users of Sendai Airport (SDJ)

Increase 2005-2000

5000 Hachinohe 1000 500 Zone 100 50 10 5000 Morioka 1000 0 100km 500 Zone 100 50 10

0 100km 500 100 50 Weekday 10 -10 2000 -50 -100 Weekday -500 2005 0 100km

Figure 2 Trip-end Distribution of the Flights at Sendai Airport (2000 and 2005) the Hachinohe zone, between 2000 and 2005.

4. CONCLUSION

In an inter-city network already constructed to certain degree of density, as in Japan, an improved link can be used connectively with other links, sometimes to a different transportation mode. For example, a newly constructed Shinkansen line can be used as middle-distance access to a local airport, and the route shift that occurs supports the air link departing at the local airport. As a result, the improvement effects are propagated in a wide, and multi-modal, way. This paper has shown quantitatively how such propagation actually occurred in Japan between 1995 and 2005. The k-th shortest path search algorithm was applied to find the set of inter-city trip routes. Remarkable changes due to the Tohoku Shinkansen expansion up to Hachinohe City in 2002 were found. This paper also analyzed the change of the service level of the air, multi-modal, and rail routes from the Hachinohe zone to other zones in Western Japan, and indicated the strong multi-modal propagations of this Shinkansen expansion on the air links at Sendai airport, proved by the increase of Sendai Airport users from the neighbor of the expanded Shinkansen line.

The most important implication of this study on transportation infrastructure planning is the Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009 need of comprehensive multi-modal planning. Historically, airway have been frequently considered as a competitive mode for railway and the substitution effects between those two modes had been gathered attentions in the planning of each mode. This study, however, shows the possibility of complementally mixture of these two different modes. Railway service can collect demands distributed thinly along the line, then can be used as a demand collector to air service from a regional airport. On the other hand, a regional airport located at the far end of a Shinkansen railway line can be used as a demand collector for the Shinkansen line to support enough frequency in the far end part of the railway network. Railway demand usually goes down along the distance from the National Capital (Tokyo in Japan), but train capacity cannot easily be decreased accordingly. Cut of train frequency cut in that portion would be harmful for the service level and result in the lost of users. Then the attractive power of regional airport at far end become helpful for affordable service. This study teaches the importance of such complementally use of the modes in transportation planning.

The methodology explained in this paper has important applications in national transportation policy formulation. In Japan, air carriers are thinking about the most appropriate service- shrinking strategy for rural airports, whilst, at the same time, they are seeking more competitive power on the trunk lines connecting the densely populated metropolitan areas. As a result, the transportation service levels for the passengers from the rural areas are not secure in the future. They will have to rely more on multi-modal routes than they do today. Thus, improvement to rail-air connectivity in rural areas may become more and more critical, and the methodology applied in this paper will be of considerable practical importance.

The future research issues to address include: the behavioral analysis on the route choice set of inter-city travelers; and the building of model describing the attributes of routes included in the choice set. Realistic policy evaluation of the proposed government projects and operational projects, such as timetable coordination and physical transfer shortening, will also be undertaken.

REFERENCES

Cascetta, E., Papola, A. (2003) A joint mode-transit service choice model incorporating the effect of regional transport service timetables, Transportation Research, Part-B, 37 , 595- 614. Givoni, M. and Banister, D. (2006) Airline and railway integration, Transport Policy , 13 , 386-397. Gonzalez-Savignat, M. (2004) Competition in air transport – the case of the high speed train-, Journal of Transport Economics and Policy , 38 , 77-108. Hazemoto, J. Tsukai, M. and Okumura, M. (2003), Evaluation of rail/air network considering alternative paths, Journal of Infrastructure Planning in Japan , 20 , 255-260. (in Japanese) Horn, M. T. (2003) An expected model and procedural framework for planning multi-modal passenger journeys, Transportation Research, Part-B, 37 , 641-660. Janic, M. (2003) High-speed rail and air passenger transport: a comparison of the operational environmental performance, Proceedings of the Institution of Mechanical Engineers, Part F, Journal of Rail and Rapid Transit, 217, 259-269. Kato, N. Ibaraki, T, and Mine, H. (1978) An O(Kn) algorithm to calculate K shortest simple Proceedings of the Eastern Asia Society for Transportation Studies, Vol.7, 2009

paths in an undirected graph with nonnegative arc length, Journal of telecommunication society of Japan , J-61A , 1199-1205. (in Japanese) Krygsman, S., Dijst, M., and Arentze, T. (2004) Multimodal public transport: an analysis of travel time elements and the interconnectivity ratio, Transport Policy , 11 , 265-275. Lythgoe, W. F., Wardman, M. (2002) Demand for rail travel to and from airports, Transportation , 29 , 125-143. Malighettia, P., Palearia, P. S. and , Redondib R. (2008) Connectivity of the European airport network: ‘‘Self-help hubbing’’ and business implications, Journal of Air Transport Management , 14, 53-65. Matsuda, Y. and Toshibe, S. (2005) Daily accessibility by air in Europe and East Asia, Proceedings of infrastructure planning in Japan , 31 , CD-ROM. (in Japanese) Miller-Hooks, E. and Patterson, S.S. (2004) On solving quickest time problems in time- dependent dynamic networks. Journal of Mathematical Modelling and Algorithms , 3, 39–71. Morichi, S. and a workshop for two layered broader regions (2005) State-of-the-art for a national planning in population decrease society; two layered broader regions as a novel land planning , Nihon–keizai–sinbunsya. Okumura, M. and Tsukai, M. (2007) Air-rail inter-modal network design, Journal of the Eastern Asia Society for Transportation Studies , 7, 180-194. O’Sullivan, P. J., Patel. T. (2004) Fragmentation in transport operations and the case for system integrity, Transport Policy , 11 , 215-225. Sato, T. and Totani, Y. (2005) Two layered borader regions and national integrated transport system, Proceedings of infrastructure planning in Japan , 31 , CD-ROM. (in Japanese) Tapiador, F. J., Burckhart, K. and Marti-Henneberg, J. (2009) Characterizing European high speed train stations using intermodal time and entropy metrics, Transportation Research , A, 43 , 197-208. Tsukai M. and Okumura M. (2003) Analysis of inter-city passenger's route choice behavior on non-shortest routes, Journal of the Eastern Asia Society for Transportation Studies , 5, 1463-1475. Tsukai M. and Okumura M. (2005) Retrospective assessment of inter-regional transportation service -- Network externality in LOS improvement --, Journal of Eastern Asia Society for Transportation Studies , 6, 328-341. Tsukai, M. and Okumura, M. (2006) Empirical analysis between round trip area (RTA) and passenger demand in Japan, ARTS (Air Transport Research Society) 2006 , Nagoya, CD- ROM 196.